Sensostat Try the demo

Textbook · Module 08

Temporal Methods: TI, TDS, TCATA

Perception dynamics over time: time-intensity, temporal dominance of sensations (TDS) with chance and significance lines, TCATA.

Module 08 of 11~20 min readinteractive + quiz
Temporal tasting of chocolate: the sequence of sensations from the first bite to the aftertaste
TDS records which sensation dominates at each moment: from "milky" at the start to "cocoa" in the aftertaste.

Taste is a process, not a photograph

Classical methods give an integral score, but perception unfolds over time: attack, plateau, aftertaste. Temporal (dynamic) methods measure exactly this dynamics.

Time-Intensity (TI) — the assessor continuously moves the intensity slider of a single attribute from the first taste to the aftertaste. Parameters are extracted from the curve: Imax (maximum), Tmax (time to maximum), AUC (area under the curve), duration, rising and falling angles. The parameters are compared by ANOVA.

TI curves: menthol cooling intensity in two chewing gums

Sample B has a lower peak but a longer plateau and a slower decline — a "mild but long" profile. An integral "menthol intensity" score would hide such a difference.

TDS: temporal dominance of sensations

TDS (Temporal Dominance of Sensations) — the assessor sees a list of attributes and at each moment marks the dominant one (the most noticeable — not necessarily the most intense). The time of each tasting is standardised to 0–100 %; dominance curves show the share of the panel for which the attribute dominates at a given moment.

The curves are read through two lines:

  • the chance level line = 1/k, where k is the number of attributes: below it, dominance is indistinguishable from random picking;
  • the significance line — the smallest share significantly exceeding chance (exact binomial test, α = 0.05); it depends on the data volume (assessors × replicates).

The TDS significance line interactive

Watch how the panel size lowers the significance line: with 10 assessors "significant" starts at 60 % dominance, with 30 — noticeably lower (≈ 43 %). Curves — milk chocolate, k = 4 attributes.

Key TDS parameters: dominance rate (the share of the panel for which the attribute dominates), time to first dominance, duration of dominance. Product comparison — difference curves with significance limits.

Time standardisation and statistical testing

Assessors taste at different tempos, so the "raw" curves are not averaged directly: the time of each run is normalised from the start of evaluation (x₀) to swallowing (x₁) — separately for the "in-mouth" phase and the "aftertaste" phase. After normalisation all curves share a common 0–100 % scale, and the panel shares at each point are binomial proportions for which the significance line is computed exactly (as in the widget above).

For the conclusion "products differ in their curves" the significance line alone is not enough — it tests individual points. The accepted approach is randomization tests (Meyners & Pineau, 2010): a curve-difference statistic (e.g. the area of the difference curve) is compared with the distribution of the same statistic under thousands of random permutations of the product labels. For TCATA data the same difference curves, confidence bands and "contrail" trajectories with uncertainty ellipses are used (Castura et al., 2016 — open access).

TDS design parameters: 8–14 assessors × 3–4 replicates (24–56 runs per product in total), 6–10 attributes in the list, mandatory training on the concept of "dominance" — this is what distinguishes TDS from intensity scales.

Case: optimising the flavour "breakdown" of milk chocolate

A panel compared three recipes (50 %, 70 %, 85 % cocoa) by TDS:

  • 50 % cocoa: the "milky" dominance holds for the first 40 % of the time, then a smooth transition to "caramel" — a harmonious profile;
  • 70 % cocoa: "cocoa" disappears too quickly, leaving an "empty" tail — consumer complaints of "lack of character";
  • 85 % cocoa: early dominance of "cocoa", late dominance of "roast"; bitterness stays significantly above the significance line for the whole second half-period.

The fix for the 70 % recipe: adding milk fat prolonged the "cocoa" dominance and removed the dip. Business effect — a 15 % growth in repeat purchases after the relaunch (a teaching case based on industry practice).

TCATA and other variants

  • TCATA (Temporal Check-All-That-Apply) — at each moment several applicable attributes can be ticked; analysis by citation curves and "fading" (sequential un-ticking).
  • MTS (multiple TDS) — parallel TDS runs by modality (taste/aroma/texture).
  • Time standardisation to 0–100 % is mandatory: assessors chew at different speeds, and without normalisation the curves cannot be combined.

Further reading

Test yourself

module 08 quiz

Three questions on temporal methods

1. How does TDS differ from TI?

TI: one attribute, continuous intensity. TDS: many attributes, discrete choice of the dominant one. The methods answer different questions.

2. In a TDS with 6 attributes, the chance level line equals…

Chance level = 1/k = 1/6 ≈ 16.7 %. Significantly above it — only dominance above the significance line (binomial test).

3. Why is time standardised to 0–100 % in TDS?

Tasting duration differs severalfold between people; time normalisation is a condition for curve comparability.

Bringing the statistics together

Module 09 — all the statistical methods of sensory science in one place: from the binomial distribution to PCA.

Module 09 →